<p>AI based automated decision-making systems (ADMs) are widely used in key decision-making processes. However, it is now well-established that such systems can discriminate against people on the basis of sensitive characteristics such as gender and race. In debates about ADMs, age is also increasingly considered as a sensitive characteristic. In this study, we argue that age is not just another sensitive characteristic, but is a special one distinguished from all sensitive characteristics due to its two unique features. First, as people get older and cross boundaries defined by age as a variable in ADMs, they are subjected to a dynamic exclusion process which causes the predictions about them to change beyond their control. This immutably mutable nature of age also redefines the impact of other sensitive characteristics such as gender, potentially with adverse consequences for disadvantaged groups. Second, the way in which age is operationalised within ADMs has an arbitrary character. This is because non-linear effects of age are usually captured by introducing age categorisations. However, different age categorisations may lead ADMs to generate substantially different outcomes for some people, without affecting the overall performance of ADMs. In this way, age categorisations can be introduced on the basis of convenience without recognising their impact on people subjected to the ADM’s outcomes. In this study we conceptually elaborate on these two unique features of age as a sensitive characteristic while also revealing them in two complementary empirical examples.</p>